A game theory based approach for distributed dynamic spectrum access

In this study, we explore the task of dynamic spectrum access based on game theory to mitigate spectrum shortage and improve network utilization in multichannel wireless networks. Usually, the available network bandwidth is limited and divided into several channels, and there exists a need for efficient reuse and adaptive allocation of such channels. During the communication process, U users compete with each other for C shared channels even without knowing accurate, complete channel state information. In order to avoid collision, traditional methods usually depend on centralized scheduling or message exchange, which are cumbersome and computationally expensive. To deal with this issue, we propose a deep Q-network, based on LSTM and fair channel allocation policy, to learn the dynamic spectrum access rules for network utility maximization. Extensive validation of the proposed approach shows that our scheme yields quite promising results.